How three Harvard and Carnegie Mellon alumni built $2.3B AI mortgage startup

The New York-based company raised $150 million in a Series D funding round on Oct. 5 from investors including Ribbit Capital and Andreessen Horowitz, lifting its valuation to $2.3 billion, according to Forbes.

Valon develops AI-powered software to automate mortgage payments, tax and insurance management, and customer service, reducing manual work and operating costs for mortgage companies.

Americans owe approximately $13 trillion in mortgages, accounting for nearly 70% of the country's consumer debt. Yet the industry responsible for collecting payments and managing loans still relies heavily on outdated technology and manual labor. The inefficiencies also affect homeowners. In 2025, the U.S. Consumer Financial Protection Bureau received more than 30,000 mortgage-related complaints, with over half involving payments and servicing.

Investors see an opportunity to transform the industry with AI, which could reduce operating costs and improve customer service.

"Think about the things we consider sexy now. Customer support is suddenly sexy. Why? Because we’ve made it AI and now everyone’s talking about it," Angela Strange, a partner at Andreessen Horowitz, told Forbes. "Mortgage is going to be the next sexy thing."

Valon Technologies' founders, (from left), Andrew Wang, Linda Du and Jon Hsu. Photos from Valon's website

From Wall Street to AI startup

The idea emerged seven years ago when Wang, a Harvard University graduate working at Soros Fund Management, became frustrated with the inefficiencies of mortgage servicing companies, which manage payments and other administrative tasks for homeowners. Despite the industry's enormous financial scale, many routine operations still depended on manual work and fragmented systems.

Wang approached his Harvard classmate Du with a proposal: she would establish a modern mortgage servicing company, and he would provide the funding. Du instead insisted they build the business together. "At the time I didn't know what mortgage servicing was," she said. "It's something the American economy runs on, but the average American off the street doesn't know what servicing is."

Before launching Valon, Wang worked at Goldman Sachs and Soros Fund Management and completed software engineering internships at Google and the White House. Du also had experience in finance and technology, including roles at Goldman Sachs, Two Sigma and Google, while Hsu studied computer science with a minor in machine learning at Carnegie Mellon University.

The three founders, now in their 30s, launched Valon in 2019 with $3.2 million in seed funding backed by Soros Fund Management. A $50 million Series A funding round led by Andreessen Horowitz followed.

In a LinkedIn post, Du said the founders saw an opportunity in an enormous industry that still depended on decades-old technology. They decided to operate a mortgage servicing business themselves because they believed firsthand experience was necessary to develop software capable of handling the industry's complex operations.

Rather than immediately selling software to established mortgage companies, the founders built their own mortgage servicing operation to demonstrate that their technology could work in an industry subject to extensive federal and state regulation.

Valon's platform automated tasks including payment tracking, tax and insurance management, and customer inquiries, replacing processes traditionally handled by employees across multiple systems. By 2025, Valon was servicing nearly 600,000 mortgages, while its estimated annual revenue more than doubled from $41 million in 2024 to $86 million in 2025. It also appeared on Forbes' Fintech 50 list for five consecutive years, while Wang and Hsu were recognized on its Under 30 list.

"Everyone in AI pitches outcomes. They talk about how more efficient their AI agents will make your operations," Wang said. "When we pitch Valon, there's a clear example that they can look at, experience, spend time with, and that's just beyond any other form of proof that you can really get."

Having demonstrated its technology through its own operations, Valon began transforming itself from a mortgage servicing provider into a software company supplying other financial institutions. In August this year, it sold its entire portfolio of approximately 800,000 serviced mortgages to Carrington Mortgage, which manages around $200 billion in loan balances.

Carrington subsequently became a customer of Valon's software platform, ValonOS. Other clients include Newrez, which services mortgages for four million homeowners, and ServiceMac, which manages approximately one million loans. Valon now employs nearly 400 people and is recruiting talent from technology companies including Google and Stripe.

AI startup limits employees' AI use

Despite building a business around AI, Valon has also imposed restrictions on how employees use the technology.

In July, Wang introduced a policy requiring most new hires, including senior recruits, to demonstrate that they understood their jobs before being allowed to use AI tools, according to Business Insider. The decision followed concerns that employees were relying on expensive AI models for simple tasks without developing the judgment needed to identify errors. "By doing the basic work, rather than relying on AI, you start to form an understanding," Wang said.

The restriction applies to most departments, although engineers are exempt because their code undergoes peer review. Wang said the policy could cut annual AI spending by up to 80%, to $4 million-$5 million, although he acknowledged in a blog post that it was "a weird decision for a company that is at the frontier of AI usage."

Valon is now using its latest funding to expand its software business, with the long-term goal of having its technology handle one in every six U.S. mortgages. The shift was part of the founders' strategy from the beginning. Rather than keeping the technology within their own mortgage servicing operation, they wanted to make it available across the industry.

"If you decide to keep the technology just for yourself and be an operating company, you don't actually change the industry," Du said. "You can probably build a $10 billion business if you had just kept the operating company and built it to its peak, but our ambitions are a lot larger than that."